Well, I think you should be able to set up some real-time monitoring that keeps 
collecting system status in the background and make some plot to observe if 
something's going wrong during the Blender workload.  An easy setup would be to 
fire up a Prometheus and Grafana docker app, and to run some node_exporters to 
collect metrics on your machine in question.  You may miss chances to reproduce 
the issue when you're looking out for them, but the monitoring system would 
catch it when you're not watching and things do go wrong.

Also wanted to say that sys-apps/dstat seems to be a good quick glance over 
what's happening in the system on different aspects in a single place.  It does 
this nice printing to the terminal once a second:

  2   1  97   0   0|   0  8192B| 292B   16k|   0     0 | 684  1662 
  2   2  96   0   0|   0     0 | 346B   17k|   0     0 | 690  1641 
--total-cpu-usage-- -dsk/total- -net/total- ---paging-- ---system--
usr sys idl wai stl| read  writ| recv  send|  in   out | int   csw 
  2   1  97   0   0|   0     0 | 560B   17k|   0     0 | 696  1630 
  4   3  93   0   0|   0     0 |1060B   34k|   0     0 |1368  3231 
  2   1  97   0   0|   0     0 |1112B   18k|   0     0 | 699  1675

It's easy to overlook things when you're juggling with multiple monitoring apps 
like htop/iotop/... while having to manage the real workload simultaneously.  
Maybe try it out when you're dealing with system activity problems next time.

Regards,
-- 
Pengcheng Xu
https://jsteward.moe

> -----Original Message-----
> From: tu...@posteo.de <tu...@posteo.de>
> Sent: Wednesday, May 6, 2020 11:10 PM
> To: gentoo-user@lists.gentoo.org
> Subject: Re: [gentoo-user] New PC hangs/lacks ?
> 
> On 05/06 04:19, tu...@posteo.de wrote:
> > On 05/06 07:07, Mark Knecht wrote:
> > > On Wed, May 6, 2020 at 3:21 AM <tu...@posteo.de> wrote:
> > > >
> > > >
> > > > Hi,
> > > >
> > > > while rendering with Blender the system performance (especially
> > > > graphic related stuff) lacks. That's not nice but it seems that
> > > > this is the way it is designed.
> > > >
> > > > What makes me a little nervous are freezes of several seconds. It
> > > > not onlu freezes but the whole graphical interface of everything
> > > > locks down (I couldn't find a corona computer virus, though).
> > > >
> > > > In the Xorg log I found this:
> > > >
> > > >     [  2808.761] (WW) NVIDIA: Wait for channel idle timed out.
> > > >
> > > > which possibly match such a moment of a freeze.
> > > >
> > > > My setup:
> > > > Blender 2.90a (alpha) and Blender 2.83 (beta) and Blender 2.82a
> > > > (stable).
> > > > All Blender versions show the same problem.
> > > >
> > > > X11/Openbox
> > > >
> > > > NVidia 484.82 as delivered by NVidia, since the Gentoo package
> > > > does not install all files of the driver which are needed for
> > > > Blender (for example to support Optix).
> > > >
> > > > No other application, which heavily uses the GPU was running at
> > > > that time.
> > > >
> > > > MSI RTX 2060 SUPER
> > > > Ryzen 5 3600
> > > > 32GB RAM
> > > > MSI Tomahawk MAX
> > > >
> > > > Does everyone has the same problems probably already solved or any
> > > > idea how I can those freezes?
> > > >
> > > > Any help or idea what causes this freezes is very appreciated! :)
> > > >
> > > > Cheers!
> > > > Meino
> > > >
> > > >
> > > >
> > >
> > > Meino,
> > >    Generically, you need to set up some sort of real-time monitoring
> > > and watch to see what is using CPU and/or I/O when the machine
> > > 'appears' to hang. I say 'appears' because the machine is probably
> > > running correctly but doing something other than Blender work.
> > >
> > >    NOTE: You didn't say that there is or isn't any disk activity
> > > when this happens.
> > >
> > >    When I look at this sort of problem I set up a second machine,
> > > ssh in with a bunch of terminals and start with 'top' and 'iotop' to
> > > watch for what process might be using resources. top watches CPU,
> > > iotop watches disk. Conceptually networking can lock up the machine
> > > but it's never happened to me.
> > >
> > >    You can also look to see if some piece of hardware is generating
> > > too many interrupts. Do
> > >
> > > watch cat /proc/interrupts
> > >
> > > in a wide terminal when not running Blender to get used to what the
> > > machine does when idle, then run Blender and see if anything is
> > > going crazy generating interrupts.
> > >
> > >    I hope some part of this helps you find your problem.
> > >
> > > Mark
> >
> >
> > Hi Mark,
> >
> > Thank you for your help and ideas.
> >
> > I will try to convince my tablet to connect to my PC and will try to
> > lock everyting which may point me into to the direction of the
> > problem.
> >
> > To complete, what I previously wrote:
> > No, the disk is not doing anything beside holidays in Honulu...so to
> > say ;)
> >
> > As soon I have something new "I'll be back"... :)
> >
> > Cheers!
> > Meino
> >
> >
> 
> Hi,
> 
> since three days these locks nags me. Now, equipped with a tablet, a ssh 
> connection
> and an eye on the interrupts...
> ...no locks happen anymore.
> 
> Grrrmpppfff...
> 
> And let me guess: In the moment when I have the least of an application for
> those locks...then they will be back... :(
> 
> Sorry, I cannot reproduce the problem anymore.
> 
> Cheers!
> Meino
> 
> 
> 

Attachment: openpgp-digital-signature.asc
Description: PGP signature

Reply via email to